It's a language that attracts casuals, but that does not mean it's incapable of being used for serious software engineering. The only scenarios where I wouldn't use python for a "serious backend thing" are scenarios in which there are dramatic cost/performance/etc consequences resulting from the overhead of using python which would be substantially reduced if using $lowLevelLanguage. Even then, there's always the option of outsourcing specific units of functionality to say, c++, anywhere the performance difference actually matters.

I would say that for the vast majority of use cases, acceptable performance could be easily achieved by simply writing better python.

Writing this reply brought to mind some absolutely atrociously inefficient ORM code I encountered in a python codebase recently. If you don't have an understanding of how to utilize SQL efficiently, it doesn't matter what language you're using to construct the SQL queries, the software engineering equivalent of warcrimes is possible in any language.